Improved Algorithms Of Median Filtering Based On Rough Set And Adaptive Theory

Image signal is often interfered by many kinds of noise when produced, transmitted and recorded. This undermines the visual effects of the image and seriously affects the following treatment like edge detection, image segmentation, etc.. Therefore, it is a very important job to use suitable methods to reduce noise (Namely carries on the image noises filter). There is a variety of image noise, where impulsive noise is one of the most common forms.At present, the commonly used methods of image filtering include linear filtering and nonlinear filtering techniques. During image processing, though the traditional linear filter can filter out noise, it often blurs image details seriously such as edges and impulse noise can not be effectively filtered. Non-linear filtering has a good immunity of the pulse signal and to a certain extent, overcomes the shortcomings of linear filters. As representatives of non-linear filtering methods, median filtering is able to effectively remove impulse noise, but it will also damage some important details of the image. In this case, many improved median filtering algorithms came into being. On the basis of these results, this article conducts an in-depth study of median filtering algorithms.The article first describes the significance and application value of studying image denoising, gives a comprehensive overview of the development of image denoising, and outlines the research focus on image denoising and trends. Then it introduces the basic theory of image denoising and conducts a detailed study of the standard median filter algorithm and its improved algorithms. In order to improve the performance of median filtering algorithm to improve the quality of image denoising, this paper does some work in the following areas:(1) Because of the complexity and strong correlation of image information, there are inevitably incomplete and uncertain problems in the image processing, which is the research content of rough set theory. Based on the research of rough set theory, a new median filtering algorithm is proposed. It tells the image pixels apart using indiscernibility relation through determining the properties of areas which the pixels belongs to, then removes noise by median filter while maintaining the same gray values of other pixels. Fuzzy membership has also been applied to determine the attributes of pixels and through this the algorithm has received good results. Experimental results show that this method has better denoising effect and details protection performance.(2) Among the many improved median filtering algorithms, adaptive median filter algorithm shows its superior performance in the process of removing impulse noise. After studying the adaptive median filter algorithm, combining the advantages of adaptive median filter in the process of removing impulse noise and to solve its problems, this paper gives an improved adaptive median filter. It detects noise according to the characteristics of impulse noise and introduces MLD to avoid miscarriage of justice between high-frequency signals and noise, then improves the estimation method of noise pixel gray value, that is filtering in accordance with the distribution of pixel gray value. Simulation results show that this method is practical and effective.(3) In the RGB space, each pixel of color images has its own R, G, B color sub-value, as long as one color component is affected by noise pollution, there will be a huge impact on the color of the pixels. Current vector methods consider the link between the three color components, but because spots polluted by noise are also involved in the formation of output value, the output value we get inevitably contains noise pollution. In order to improve the denoising results, the improved adaptive median filter algorithm is applied to color image denoising areas, R, G and B color component is processed respectively.In order to verify the effectiveness of the algorithm, this paper adopts VC++6.0 programming environment to test the images with different noise intensity respectively. The experimental results are compared with traditional denoising methods. The results show that the proposed algorithm can effectively remove the noise and retain image detail to some extent. Especially in the case of the high noise density, it has greater superiority than any other filtering algorithm.